Internet Survey Method in the 2010 Census and Challenges to the 2015 Census in Japan

Internet Survey Method in the 2010 Census and Challenges tothe 2015 Censusin Japan Population Census Division, Statistics Bureau of Japan Hideki Koizumi November, 2012

Summary • The Population Census of Japan is conducted every five years --- Next Census will be conducted in 2015. • At 2010 Census, we introduced internet method for the first time, all of Tokyo area --- internet response rate was 8.3% (very low). • Problems: changing environments --- ● Increasing One-person households, ● Complicating survey method (handing over, mail and internet), --- data quality might be influenced such causes. • Challenges to the Next Census --- ● Introduce Step by Step method to increase internet response rate --- automatically checking will be done (23.7% in 1st exam survey) ● Available to response by Smartphone --- for one-person households ● Building Response Management System

Contents • History • 2010 Census Methodology • 1st Exam Survey of the 2015 Census • Lessons Learned

History • 2010 Census Methodology • 1st Exam Survey of the 2015 Census • Lessons Learned

History • The Population Census of Japan is conducted every five years (excluding 1945 because of WW2). • First: 1920, Latest: 2010(19th) • Common feature through all Censuses before 2005 : ● Taken as of 0:00 a.m., October 1. ● Used One questionnaire form for simplicity (with about 20 items) --- no “long and short” etc. ● Distributing and Collecting by enumerators ● Self-enumeration (background: high literacyrate since 19th century)

History • Common feature through all Censuses before 2005(cont’d) : ● Conducted through the channels:

History • But, changing environments ---“Increasing one-person households” indicates “Increasing households that they are not at home in daytime”. Numbers of Households by Size (1990-2010)

2010 Census Methodology • Dealing with Absence of household members, we introduced --- ● Collecting by Mail (All area) ● Internet response method (only Tokyo) • 2010 Census Methodology: • ● Distributing questionnaire --- by Enumerators • ● Collecting --- Handing over, Mail and Internet (only Tokyo)

2010 Census Methodology • Survey items are the following 20 items: [for household members(15)] (1)Name (2)Sex (3)Year and month of birth (4)Relationship to the household head (5) Marital status (6) Nationality (7) Duration of residency at the current domicile (8) Place of 5 years previous residence (9) Education (10) Type of activity (11) Name of establishment and kind of business(Industry) (12) Kind of work(Occupation) (13) Employments status (14) Place of work or location of school (15) Transportation to the place of work or the location of school [for households(5)] (1) Type of household (2) Number of household members (3) Type and tenure of dwelling (4) Area of floor space of dwelling (5) Type of building and number of stories

2010 Census Methodology • Questionnaire form (size: A4 paper) --- For 4 persons (at most) --- OCR - Recognized

2010 Census Methodology • Flowchart

2010 Census Methodology • Picture of form of internet (1) [Number of household members] Keying Male number and Female number, then total number is automatically entered.

2010 Census Methodology • Picture of form of internet (2) [Year and month of birth] If keying by mistake, error message appears. (this example is the case of range error.)

2010 Census Methodology • Response by Internet --- Finally, 529,154 responses (8.3% of Tokyo). Extension

2010 Census Methodology • Internet response rate was lower than other countries. (mail: about 60%, enumerators: about 40% in Japan). --- for the Next 2015 Census, and for the cost effective, we have to increase internet response rate more than 2010. • It appears “Starting Day Effect”, “Ending Day Effect” and “Holiday Effect”. --- These effects are positive, but in some other countries, “Holiday Effect” is negative. • To increase number of holidays in internet period is important for us, and also we have to try other methods.

1st Exam Survey of the 2015 Census • The First Examination Survey of the 2015 Population Census was conducted in June 2012: ● Voluntary survey (approved survey from statistical standards section, but no obligation) ● Sample survey (sample size was about 19,000 households in 14 cities) ● With two internet methods (differences: internet response period)

1st Exam Survey of the 2015 Census • Two methods were following: ● Step by Step method (similar to Korean method) --- Former internet response period with no questionnaire and after other response(handing over and mail) period ● Parallel method (same as 2010 Census) --- Distributing questionnaire with internet ID, and households can choose three methods (handing over, mail and internet) in the same period.

1st Exam Survey of the 2015 Census • Step by Step method Internet Response period (10 days) Handing over and mail period Specify non-response via internet Distributing questionnaire to non-response households Non-response Follow-up Distributing internet ID to all households (with no questionnaire) • Parallel method Internet Response period (27 days) Handing over and mail period Non-response Follow-up Distributing questionnaire with internet ID to all households

1st Exam Survey of the 2015 Census • 1st exam survey’s internet response rates (preliminary results) were: ● Step by Step method : 23.7% (in spite of only 10 days) --- Automatically lockable apartments : 33.8% --- Efficiency apartments : 13.3% --- Houses : 27.8% --- Other apartments, etc. : 19.8% ● Parallel method : 6.0%

1st Exam Survey of the 2015 Census • Step by Step method’s rate was 4 times Parallel method’s rate. --- We guess many households can answer both “internet” and “mail”. • To make the period which households are restricted their response method, we can lead these households to response by internet. • “Starting Day Effect”, “Ending Day Effect” and “Holiday Effect” were appeared as well as 2010 Census.

1st Exam Survey of the 2015 Census • Why rate of Efficiency apartments was low? --- We think “One-person household with no desktop or laptop computer” is increasing because Smartphone has a role of computer and efficiency apartments in Japan are very small (maybe not afford to put computers). • We think “Available to response by Smartphone” can increase internet response rate of Efficiency apartments.

Lessons Learned • Introducing internet method has --- [Advantages] ● Reducing respondent burden ● Data quality is high --- automatically checking ● Reducing period of processing data (data entry, data check and so on) [Disadvantages] ● Complicating survey method --- Increasing burden for clerks and enumerators ● Increasing duplication

Lessons Learned • “Cost Effective” is both Advantage and Disadvantage. --- If response rate is high, then it becomes advantage. Cost Mail Internet Fixed cost of internet facilities Response rate

Lessons Learned • There are risks : ● Increasing One-person households, ● Complicating survey method (handing over, mail and internet), • So, we are planning to increase internet response rate and to conduct survey more efficiency : ● Available to response by Smartphone ● Building Response Management System to specify non-response via internet more efficiency, and to know response situation precisely

Lessons Learned • For the Next 2015 Census, we are also planning 2nd Examination Survey of the 2015 Population Census. Internet Response Rate by Survey

Thank you for your attention! h2.koizumi@soumu.go.jp

Internet Survey Method in the 2010 Census and Challenges to the 2015 Census in Japan